57 lines
1.4 KiB
Markdown
57 lines
1.4 KiB
Markdown
|
---
|
||
|
|
base_model: Qwen/Qwen3-4B-Instruct-2507
|
||
|
|
datasets:
|
||
|
|
- alfworld
|
||
|
|
- dbbench
|
||
|
|
language:
|
||
|
|
- en
|
||
|
|
license: apache-2.0
|
||
|
|
library_name: HuggingFace Transformers
|
||
|
|
pipeline_tag: text-generation
|
||
|
|
tags:
|
||
|
|
- agent
|
||
|
|
- tool-use
|
||
|
|
- alfworld
|
||
|
|
- dbbench
|
||
|
|
---
|
||
|
|
|
||
|
|
# Qwen3-4B-Instruct-2507-LoRA-AgentBench
|
||
|
|
|
||
|
|
This repository provides a RL'ed model fine-tuned from
|
||
|
|
**Qwen/Qwen3-4B-Instruct-2507** using HuggingFace Transformers.
|
||
|
|
|
||
|
|
## Training Objective
|
||
|
|
|
||
|
|
This model is trained to improve **multi-turn agent task performance**
|
||
|
|
on ALFWorld (household tasks) and DBBench (database operations).
|
||
|
|
|
||
|
|
## Training Configuration
|
||
|
|
- Base model: Qwen/Qwen3-4B-Instruct-2507
|
||
|
|
- Method: Agentic Reinforcement Learning
|
||
|
|
- Max sequence length: 8192
|
||
|
|
- Learning rate: 1e-06
|
||
|
|
|
||
|
|
## Usage
|
||
|
|
|
||
|
|
```python
|
||
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer
|
||
|
|
from peft import PeftModel
|
||
|
|
import torch
|
||
|
|
|
||
|
|
base = "fujiki/qw3-4b-v17-gs180"
|
||
|
|
|
||
|
|
tokenizer = AutoTokenizer.from_pretrained(base)
|
||
|
|
model = AutoModelForCausalLM.from_pretrained(
|
||
|
|
base,
|
||
|
|
torch_dtype=torch.float16,
|
||
|
|
device_map="auto",
|
||
|
|
)
|
||
|
|
```
|
||
|
|
|
||
|
|
## Sources & Terms (IMPORTANT)
|
||
|
|
|
||
|
|
Training data: u-10bei/sft_alfworld_trajectory_dataset_v5, u-10bei/dbbench_sft_dataset_react_v4 and task enviromnents for the Agentic RL of LLM.
|
||
|
|
|
||
|
|
Dataset License: MIT License. This dataset is used and distributed under the terms of the MIT License.
|
||
|
|
Compliance: Users must comply with the MIT license (including copyright notice) and the base model's original terms of use.
|